Talk:Nationwide opinion polling for the 2020 United States presidential election/Archive 1

Archive 1

Explanatory notes on formatting choices

(I'll fill this out in the future with more extensive explanation of choices and things that may be potentially changed.)

• Overall formatting
  • Links and citations
  • Overall table formatting
• Aggregation
• Graphs
• Lead columns
• Sortability
• Discrete candidate listings
• LV/RV (historical, may want to change)
• Pollster clients
• Partisan pollsters and clients
• Maps, margins of error, and designation of states' status based on polling
• Eventual split

Mélencron (talk) 03:09, 7 January 2019 (UTC)

Order of candidates

Alphabetical order doesn't seem suitable here, with insignificant candidates strewn all over the article, while some major candidates are down the bottom. Wondering if anyone has any alternatives. Onetwothreeip (talk) 23:37, 1 March 2019 (UTC)

It's the most neutral/consensus order, which I preserved when I split this list from the original article (referring to 2020 United States presidential election, not the combined national + state article). See Talk:2020 United States presidential election/Archive 5#How should we sort the polls? as well as Talk:2020 United States presidential election/Archive 5#Polls in Alphabetical Order? for past discussions on the topic – it was the status quo order at the time and I didn't change it. I'm open to hearing alternate proposals (one was by listing most-to-least-polled in descending order, which seems half-sensible and yields, by my count, the following order: Biden/Warren 27, Sanders 24, Harris 19, Booker 17, Gillibrand 10, O'Rourke 4, Bloomberg/Delaney/Holder/Klobuchar 2, Brown/Bullock 1), but I'm not sure if they can be deemed entirely neutral. (You could also determine it based on how Democrats are currently polling in the primaries, but I'm also not sure if that's an entirely neutral criterion here, even if I think it would be more useful and navigable than the current order.) Since this was a split from 2020 United States presidential election, it might also be useful to crosspost/post a notice of this discussion on Talk:2020 United States presidential election as well, as I'm not sure if I have any better ideas (and most editors there don't have this page watchlisted). Mélencron (talk) 01:36, 2 March 2019 (UTC)

I think ordering them by the amount of polls is the best option. This best reflects the reliable sources too, since they are more concerned with and prioritise candidates like Biden and Sanders over those minor candidates with very sporadic polling. Onetwothreeip (talk) 02:45, 2 March 2019 (UTC)

Done. Mélencron (talk) 02:51, 2 March 2019 (UTC)

Thanks, looks like a reasonable and informative sorting criterion. — JFG ^talk 20:06, 2 April 2019 (UTC)

Specific Other Candidates

If a poll were to include a specific 'other' candidate (in this case Justin Amash), should a whole new column be added for that candidate, or would they just be listed under the 'other' column?74.110.151.202 (talk) 16:34, 14 August 2019 (UTC)

The "Other" column is sufficient at this point. The addition of a new column should be reserved for candidates that receive significant results in multiple polls. --Spiffy sperry (talk) 19:41, 14 August 2019 (UTC)

OK, thanks. Assuming significant results means 10% or plus?74.110.151.202 (talk) 12:57, 17 August 2019 (UTC)

Request for chart

Kinda like the following examples: 1, 2, 3, etc. — Preceding unsigned comment added by Mapmaker345 (talk • contribs) 17:41, 17 August 2019 (UTC)

Chart Request

Per above, but since no one responded here's a new heading, now that its going to be Trump v. Biden, perhaps we should get a chart. --TDKR Chicago 101 (talk) 21:37, 8 April 2020 (UTC)

Date Format

I suggest that the dates format be changed to one rationally table sortable. Deancarmeli (talk) 13:23, 16 April 2020 (UTC)

@Deancarmeli: I don't know a date format with ranges which is sortable without adding sort keys per Help:Sorting#Specifying a sort key for a cell. PrimeHunter (talk) 18:39, 16 April 2020 (UTC)

@PrimeHunter: We can use formats such as {{Date table sorting|2020-04-17}} or {{Date table sorting|April 17, 2020}} that yields April 17, 2020.
Ranges will be displayed with the full dates, not abbreviated. April 14–16, 2020 will be changed to {{Date table sorting|2020-04–14}} — {{Date table sorting|2020-04-16}} and seen as April 14, 2020 — April 16, 2020.

If we're at it, I'll also recommend a shift to a more "search-friendly" date format such as DD-MM-YYYY, like {{Date table sorting|April 17, 2020|format=dmy}} that yields 17 April 2020

@Deancarmeli: Two full dates looks clumsy for a brief range in a table cell, and it's against MOS:DATERANGE. And if you write that much code then you may as well just add a sort key. I don't expect readers to search by date. The tables are sorted by date and it's easy to just scroll to a period of interest. The American date format mdy is preferred per MOS:DATETIES. PrimeHunter (talk) 12:15, 17 April 2020 (UTC)

JHK Forecasts

Would it be a good idea to add JHK Forecasts, an increasingly popular forecast, to the polling aggregate? https://projects.jhkforecasts.com/presidential-forecast/ Spinosaurus75 (Dinosaur Fan) (talk) 13:46, 13 July 2020 (UTC)

Since there's no response, I'll just add it I suppose. Spinosaurus75 (Dinosaur Fan) (talk) 01:15, 14 July 2020 (UTC)

Looking at their methodology page, they do not seem to be independently aggregating polls, just copying from 538, complete with using 538's pollster ratings for determining poll weights. They then try to model drift toward "fundamentals" and toss in the assessment of various experts (as an added bit of herding).

I don't know if there is some standard that has previously been used for deciding which aggregates should be considered for inclusion; on the 2016 polling page, for example, the New York Times was included as a separate aggregator even though it was evidently using the same poll list (with different weights) as the Huffington Post. However, with JHK reusing both 538's polls and their pollster ratings, in addition to its structure as a forecast rather than an aggregation, I'm disinclined to include it.

If their polling weights, despite using 538's pollster ratings, are sufficiently different from 538 to produce a noticeably independent estimate of the state of the race (a cause not aided by their decision to intentionally herd toward existing expert ratings), I would be more inclined to include them.Gambling8nt (talk) 04:08, 14 July 2020 (UTC)

They use certain elements of FiveThirtyEight but not necessarily copying their polling aggregates. In fact they published their forecast before FiveThirtyEight published their polling aggregates. If other stuff such as fundamentals and expert ratings shouldn't be used, I'm fine with simply using the polling part of their forecast, but I honestly don't mind either way, since the other aggregators adjust the results with certain methods anyway, especially FiveThirtyEight. Using their full forecast result isn't necessarily a bad thing. Spinosaurus75 (Dinosaur Fan) (talk) 05:51, 14 July 2020 (UTC)

Splitting pollsters that measure 3-way, 4-way, or 5-way race

Is it really necessary in this election to split these polls? 3.5 months from Election Day only 3 pollsters (Change Research h, Researxh Co., Redfield have decided to poll minor candidates. This is not 1992: /1996, or even 2000/2016 where minor candidates have any visibility.

Is it really worth separating a few pollsters from the chart because they prompted for parties getting 2%? Instead of just having one global chart and “other” can always have a link note to show what % “other” breaks out to? — Preceding unsigned comment added by Mikemikem (talk • contribs) 14:08, 15 July 2020 (UTC)

Democracy Institute/Express Newspaper polling dispute

I can see by the edit history that there is a dispute over the inclusion of polls by the Democracy Institute/Express Newspaper. As an uninvolved editor, can those involved please explain what is going on? --Minoa (talk) 10:55, 5 August 2020 (UTC)

I am only indirectly involved. I frequently update this polling page, so I deleted 2 of the 3 polls after 1 of the 3 had been deleted, solely for consistency. It does not make sense to have one or two out of three polls included from the same pollster. It needs to be zero or all three.

Tophat John and Sociallyobservant4297 are going back and forth deleting.

It appears the argument for inclusion is that it is a poll published in a major newspaper (albeit a tabloid UK paper known for not always being above board). The argument for exclusion is that none of the major US polling websites (mainly looking at 538, which literally includes every poll, including campaign-internals) does not include this poll.

Mikemikem (talk) 13:27, 5 August 2020 (UTC)Mikemikem

I'm leaving it up as long as the dispute is unresolved and have no problem continuing to do so, but it's not just left unincluded by 538; it's banned by them because it's suspected of faking data. With exceptions like the work of pollsters under review, it's my understanding that we don't usually leave such polls up here. --PutItOnAMap (talk) 11:57, 30 August 2020 (UTC)

Changing chart y-axis to start at zero

I'm changing the y-axis to start at zero, and just wanted to explain here. I accept that it's open to debate whether line charts' y-axes should always start at zero. But I think a voter-split projection chart is a clear case where it should. For many casual viewers, their first impression (until they look closely at the y-axis labels – which many won't) would be that Biden is on course to win about three times as many votes as Trump, giving the false impression that there's almost no contest! In fact, Biden only leads Trump by about 10 percentage points (about 50% vs. 40%). Starting the y-axis at zero makes this clearer. Denbosch (talk) 13:14, 14 August 2020 (UTC)

I agree, but let's make the top 100% as well, why do we have it at 55% (it makes the difference look 2 times larger, than it would look on 100% scale)? --195.60.68.148 (talk) 15:26, 14 August 2020 (UTC)

Disagree. The main reason this chart is informative is to visualise the change in polling over time, which is much harder to interpret with this scaling. If there's a risk of the chart's scaling being misinterpreted, the better solution is to add an explicit warning about it above/below the chart to draw attention, in my view. I won't revert the change myself, though, in case others disagree. 2A00:23C7:3700:1901:33CC:7F69:3FFD:986F (talk) 13:32, 15 August 2020 (UTC)

I am in agreement with both of these points. Either use 0 to 100% as the scale, or go back to the original scale, in order to highlight the difference. — Preceding unsigned comment added by 142.116.93.126 (talk) 15:14, 7 September 2020 (UTC)

Adding 4-way aggregation poll

Should we add it since there's now a 4-way (third parties included) such as on 270toWin (scroll down to the 4-way version) & RCL? — Preceding unsigned comment added by Syaz351 (talk • contribs) 06:41, 11 September 2020 (UTC)